diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
index 5c6890db82fd..701e2e7760d8 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
@@ -12,7 +12,6 @@
 from ...schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
 from ...utils import logging
 from . import StableDiffusionPipelineOutput
-from .safety_checker import StableDiffusionSafetyChecker
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -53,7 +52,6 @@ def __init__(
         tokenizer: CLIPTokenizer,
         unet: UNet2DConditionModel,
         scheduler: Union[DDIMScheduler, PNDMScheduler, LMSDiscreteScheduler],
-        safety_checker: StableDiffusionSafetyChecker,
         feature_extractor: CLIPFeatureExtractor,
     ):
         super().__init__()
@@ -78,7 +76,6 @@ def __init__(
             tokenizer=tokenizer,
             unet=unet,
             scheduler=scheduler,
-            safety_checker=safety_checker,
             feature_extractor=feature_extractor,
         )
 
@@ -284,15 +281,11 @@ def __call__(
         image = (image / 2 + 0.5).clamp(0, 1)
         image = image.cpu().permute(0, 2, 3, 1).numpy()
 
-        # run safety checker
-        safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
-        image, has_nsfw_concept = self.safety_checker(
-            images=image, clip_input=safety_checker_input.pixel_values.to(text_embeddings.dtype)
-        )
-
         if output_type == "pil":
             image = self.numpy_to_pil(image)
 
+        has_nsfw_concept=False
+
         if not return_dict:
             return (image, has_nsfw_concept)